Google has unveiled a long-term research project called Suncatcher, which aims to move some machine learning operations into space. Google CEO Sundar Pichai announced it in an interview with host Shannon Bream on Fox News Sunday. According to him, the company plans to begin sending small racks (servers) into space aboard satellites in 2027, where they will be tested and then gradually expanded. The idea stems from the need to harness solar energy, which is available in enormous quantities in space—as much as one hundred trillion times more than is produced on Earth. Pichai mentioned that within ten years, having data centers beyond Earth could become commonplace, helping to reduce the strain on terrestrial resources. The project is part of Google's effort to pursue ambitious goals, such as this "moonshot," which is intended to address the growing environmental problems associated with artificial intelligence.
In the "Google AI: Release Notes" podcast, Pichai added that they hope to have a TPU in space in 2027, which is Google's own chip for artificial intelligence. This means that a small device could operate in orbit and use solar energy directly, without the losses that occur on Earth. In a press release, Sally Radwan of the United Nations Environment Programme warned that the environmental impact of artificial intelligence is concerning because it involves the extraction of rare materials, electronic waste, water consumption for cooling, and greenhouse gas emissions from operating data centers.
Growing Energy Demand Due to Data Centers
In the US, data centers are dramatically increasing electricity consumption, which, according to a 2025 report by McKinsey & Company, is unique on a global scale. This report, titled Global Energy Perspective 2025, estimates that average annual growth in energy demand for data centers will reach 17% globally between 2022 and 2030. In the US, it is even faster—up to 25% annually. By 2030, data centers could consume more than 14% of the country's total electricity, more than triple the level in 2023.
This boom is linked to the construction of enormous data center campuses by major technology companies. For example, the Stargate campus is being built in Texas with an Oracle data center that will host OpenAI servers and be powered by its own natural gas power plant. In Virginia, Arizona, and Louisiana, energy companies such as Entergy Louisiana are seeking approval to build new power plants and transmission lines worth billions of dollars. In August, regulators approved a plan allowing Entergy Louisiana to recover $5 billion (approximately CZK 115 billion) for the construction of three new natural gas power plants for Meta's data center. The McKinsey report warns that because of this demand, fossil fuels will remain a significant part of the energy mix until 2050, as technologies such as carbon capture and hydrogen energy are developing more slowly than expected.
Industrial development could double global electricity demand by 2050, with data centers being a key factor in Western Europe, China, and North America. In the US, this means that energy companies expect to add 60 gigawatts of new capacity solely for data centers by the end of the decade—enough to power six large cities.
Air Pollution Caused by AI
Artificial intelligence also brings serious air pollution problems, as shown by a study from the University of California, Riverside, and the California Institute of Technology. This paper, titled "The Unpaid Toll: Quantifying the Public Health Impact of AI," estimates that emissions from data centers could cause as many as 1,300 premature deaths annually in the US by 2030—a 36% increase compared with the current number of asthma-related deaths. Researchers Shaolei Ren and Adam Wierman examined emissions of nitrogen, sulfur, and fine particles with a diameter of 2.5 micrometers or less, which penetrate deep into the lungs.
For example, training a single large language model, such as Meta's Llama 3.1, would generate as much pollution as a car traveling from New York to Los Angeles and back 10,000 times, for a total of nearly 40 million kilometers. The total health costs associated with artificial intelligence could reach $20 billion (approximately CZK 460 billion) within six years. Last year, they were estimated at $5.6 billion (about CZK 129 billion). The study predicts that data centers will consume 11% to 12% of total electricity in the US by 2030, up from 3% to 4% last year.
Diesel generators used as backups in data centers are particularly dangerous. In Virginia, which has one of the densest concentrations of data centers in the world, these generators could cause 13 to 19 additional deaths annually when operating at 10% of permitted emissions, or as many as 130 to 190 at full utilization. This would result in annual health costs of $220 million to $300 million (CZK 5 billion to CZK 7 billion) under the lower scenario, or up to $3 billion (CZK 69 billion) under the higher one. The researchers found that pollution disproportionately affects economically disadvantaged communities and recommend greater transparency from companies such as Amazon, Google, Microsoft, and Meta, which do not disclose air pollutant emissions in their sustainability reports.
These three areas—Google's space project, growing energy demand in the US, and air pollution—point to a common problem: artificial intelligence consumes enormous amounts of energy, placing a strain on Earth. Google is trying to move some of this burden into space, where solar energy is not limited by weather or the cycle of day and night. Google is searching for solutions, but for now it is clear that the growth of artificial intelligence brings challenges that affect everyone's health and the environment.
Sources: businessinsider.com, businessinsider.com and businessinsider.com
